An Improved Method for the Elderly Fall Recognition System Based on YOLOV7*
Ruifeng Xu, Yuyuan Mai, Junlong Chen, Zhen Zhang · 2023
With 761 million people aged 65 and over in 2021, the global population is now aging more and more severely. Falls are the leading cause of injury-related death for people over 65 years old. To ensure the safety of the elderly, our team has previously developed a fall warning system for elderly people living alone, but the average judgment accuracy rate is only 87.68%. In order to reduce the occurrence of judgment errors as much as possible and improve the detection performance of the system, this paper improves the accuracy of detection by reasonably reducing the fall judgment score and introducing the YOLOV7 for further identification of possible falling posture. Compared with the previous system, the accuracy of the improved system increased from 87.68% to 97.41%. Consequently, the improved method proposed in this paper has greater detection performance, robustness and lower error rate.